Download bank credit and the housing market

1
BANK CREDIT AND THE HOUSING MARKET
IN OECD COUNTRIES
Philip Arestis, University of Cambridge, UK, and University of the
Basque Country, Spain
Ana Rosa González, University of the Basque Country, Spain
Abstract
Relevant economic literature frequently focuses on the impact of credit shocks on
housing prices. The macroeconomic doctrine of the ‘New Consensus Macroeconomics’
completely ignores bank credit. The ‘great recession’, though, has highlighted the
importance of bank credit. The purpose of this contribution is to re-visit this important
macroeconomic variable. Consequently, we propose to endogenise the volume of bank
credit by paying special attention to those variables that are related to the real estate
market, which can be considered as key to the evolution of bank credit. Our theoretical
hypothesis is tested by means of a sample of 9 OECD economies from 1970 to 2011.
We apply the cointegration technique for the latter purpose, which permits the
modelling of the long-run equilibrium relationship and the dynamics of the short run
along with an error-correction term.
Keywords: bank credit, collateral channel, housing market, OECD countries, empirical
modelling.
JEL Classification: C22, R31.
2
1. Introduction
The events leading to the August 2007 financial crisis have shown that a private debt
crisis can provoke a systemic banking crisis. This contribution proposes to examine the
possibility of endogenous bank credit and consider the housing market as the main
source of demand for credit in the private sector. In particular, our approach assumes
that agents demand bank credit that they need to satisfy their economic activities. In this
context, the final volume of credit is not a supply-led variable; it is rather a demanddetermined variable in the sense that the banking sector provides credit in response to
the demand for it. For the purposes of this contribution, we focus on the demand for
credit, which emanates essentially from mortgages to acquire housing.
After endogenising bank credit to the domestic sector and discussing the main
determinants of our theoretical framework, we proceed to investigate this proposition
empirically. We employ for this purpose cointegration and error-correction techniques
in a sample of 9 OECD economies over the period 1970 to 2011.
The remainder of this paper is structured as follows. In section 2 we put forward
our theoretical framework. In section 3 we explain our approach and results of our
empirical investigation: sub-section 3.1 presents the econometric technique, sub-section
3.2 discusses the data sources, and sub-section 3.3 reports the empirical results of the
long-run equilibrium relationships and the short-run dynamics. Section 4 offers an
overall discussion of the theoretical and empirical results. Finally, section 5 summarises
and concludes.
2. Bank Credit
We begin with the traditional notion of the ‘dynamic monetized production economy’,
which considers that any flow of production needs a previous flow of credit. In this
sense businessmen determine the level of production, prices, dividends and salaries
based on their expectations of future demand. The development of the production
process requires credit, which is the element that drives and permits the evolution of the
economy. The only role in this process, which has to be played by the banking sector, is
to provide credit in response to businessmen’s demand. In this theoretical approach, the
Central Bank influences the cost of credit through manipulating the interest rate under
its control. The Central Bank also affects credit standards, which ensure the proper
functioning of the economy, through generating the volume of reserves that commercial
banks require in order to provide the credit that is demanded.
3
We develop our model by considering a close economy without public or
foreign sectors where the following agents interact: households, firms, and the banking
system. 1 We also adopt the following assumptions: (a) there are two factors of
production; labour, which is remunerated by means of wages, and capital whose
remuneration is the profits obtained from the production process; (b) firms develop
productive activities in order to provide those goods and services as required; (c) firms
are grouped into two categories: real estate firms, which include property developers,
and non-real estate companies; the latter comprise of the rest of the firms in our model;
(d) all the profits are saved; (e) households use their income to consume goods and
services; (f) the acquisition of housing assets is the main investment decision
undertaken by households, since it is normally financed by getting into debt and
comprises of an important part of the income earned during households’ working life;
and (g) the Central Bank sets the short-term interest rate and the credit standards that the
commercial banks have to face in order to provide the required credit.
We modify the original framework of the ‘monetary production economy’ by
accounting for the housing market and the creation of related credit in order to show
how those factors, which determine the equilibrium in the housing market, explain the
evolution of credit in the private sector. 2 After the distribution of the income generated
in the production process to remunerate households, a share of initial credits borrowed
by firms comes back to the banking sector as deposits, which constitutes banking sector
reserves. All the households use their current income to consume goods and services.
Some of them decide to purchase a property, which means borrowing external resources
from the banking sector in order to face its cost. The desire to invest in real estate assets,
i.e. demand for housing, is mainly a function of housing prices, disposable income and
the mortgages rate, since these elements determine the affordability of dwelling, which
is defined as price-to-income ratio. As a result, households’ demand for housing is
summarised as in equation (1):
DH = DH ( PH , RDY , MR)
-
1
2
+
-
The banking system includes the Central Bank and commercial banks.
See Graziani (2003) for an explanation of the original ‘monetary production economy’ theory.
(1)
4
where the demand for housing, DH, is a positive function of real disposable income,
RDY, and a negative one with respect to housing prices, PH, and the mortgage rate, MR.
The sign below a variable indicates the partial derivate of DH with respect to that
variable.
To account fully for the dynamics in the housing market and include the role of
house developers, i.e. firms, we hypothesise the supply for real estate assets as in
equation (2):
S H = S H ( PH , RRI )
+
(2)
+
which shows how the supply of housing, SH, is positively influenced by housing prices,
PH, and real residential investment, RRI.
Equations (1) and (2) can be set equal to each other at equilibrium, and solving
the resulting equality, we obtain the quantity and the price of housing as shown in
equation (3) and (4) respectively:
PH = PH ( RDY , MR, RRI )
+
-
-
QH = QH ( PH , RDY , MR)
+
+
(3)
(4)
-
where the variables are as in equations (1) and (2).
Those households who decide to participate in the housing market are compelled
to get into debt. 3 In this context, the commercial banks face households’ demand for
credit, select those borrowers who are credit-worthy and lend part of their reserves in
the form of mortgages, which are required to purchase dwellings. As a result, the
3
There is, of course, another group of households who invested in real estate assets in the past and have to
employ a fraction of their income to repay their mortgages.
5
demand for credit, which is required to acquire those properties that households desire
depends on the equilibrium in the housing market as displayed in equation (5):
DC = DC ( PH (⋅), QH (⋅))
+
(5)
+
where the symbols have the same meaning as in previous equations, except DC, which
stands for credit demand.
Equation (6) is obtained by substituting equations (3) and (4) in equation (5),
which provides an extended version of the credit demand equation:
DC = DC ( PH , RDY , RRI , MR)
+
+
+
(6)
-
where the variables are defined as in equation (3), (4) and (5).
The volume of bank credit, which is generated in this way is given by the
interaction between a horizontal supply of credit, at a given level of interest rate, and the
demand for credit, which satisfies the credit standards as determined by the central
bank. 4 The supply of credit, which is included in the model, is defined as in equation
(7):
S C = S C (MR)
(7)
where Sc is the supply of credit, and MR, the mortgage rate.
At equilibrium, demand for and supply of credit are equal, which permits us to
put equations (6) and (7) equal to each other and solve the resulting equality for bank
credit, BC, at the steady state of our model; as in equation (8):
BC = BC ( PH , RDY , RRI , MR)
+
4
+
+
(8)
-
The assumption of a horizontal supply of credit is the reflection of the notion of a horizontal supply of
money (Kaldor, 1982, Moore, 1983). In the context of endogenous money, the central bank determines
the rate of interest at which it provides the liquidity that is required by the credit-worthy demand. The
stock of money-credit is driven by the demand for money and is out of the control of the banking
institutions (see also Lavoie, 1984, for further analysis).
6
which displays how the volume of bank credit, BC, is positively related to housing
prices, PH, real disposable income per capita, RDY, and real residential investment, RRI.
The volume of bank credit is also a negative function of the mortgage rate, MR. This
shows how the volume of bank credit is explained by those variables, which define the
affordability of the properties, i.e. housing prices, disposable income and the mortgage
rate; and the flow of new real estate assets, which is measured by real residential
investments. 5
Our approach is different from the traditional one of the housing literature where
the major role in the housing market is given to the supply of credit. 6 Our model, which
is rooted in the Banking School doctrine, considers the supply of credit as a residual
element, since credit is demand-driven and the supply of money-credit is horizontal at
the level of the interest rate as determined by the Central Bank. As a result, the moneycredit stock, i.e. the volume of bank credit, is determined by demand, which comes from
the private sector, since commercial banks would provide all the volume of credit,
which satisfies the credit standards as determined by the Central Bank. 7
The first determinant of banking credit included in equation (8) is housing
prices, PH. 8 Its relevance is twofold: (a) housing prices are key in the determination of
the affordability of dwelling; and (b) its evolution exerts important effects on the
dynamics of the credit market and the housing market via the ‘collateral’ channel. 9 In
particular, the ‘collateral’ channel captures how an increase in housing prices introduces
relaxation of the conditions that a borrower has to face in order to obtain a mortgage.
This is so since the ‘protection’ against a possible borrower’s default, which is offered
to the lender, is stronger due to the fact that the asset that is collateralised has a higher
5
The model assumes that new properties are acquired by households who do not have a previous
residential asset, which can be sold in order to finance the new purchase. Demand for credit has to
account explicitly for residential investment since the final volume of bank credit is a proportion of the
value of the flow of new dwellings, which are produced in response to the demand for housing.
6
See Adelino et al. (2012) where empirical evidence is provided to support the assumption that easy
access to credit fuels housing prices.
7
See Daugherty (1942) for further details on the Banking School approach.
8
See Goodhart (1995) and Hofman (2004) for empirical evidence of the positive impact of housing prices
on credit.
9
The ‘collateral’ channel is considered as a variant of the ‘financial accelerator’ introduced by Bernanke
et al. (1999). According to this idea, under the presence of asymmetric information those economic agents
who opt for borrowing resources, instead of using their own resources, in order to undertake their
projects, have to face a prime. This moves procyclically due to two alternative factors: either procyclical
movements in the value of those assets that can be used to secure the mortgage (Kiyotaki and Moore,
1997); or countercyclical changes in terms of the cost of information between external and internal source
of finance (Bernanke and Gertler, 1989). The cost that investors have to pay for these resources depends
on their net worth and exerts a procyclical impact on their capacity to obtain finance. Under these
circumstances the credit market amplifies the fluctuations in production and other macroeconomic
aggregates of the real economy (see also Bernanke, 2007).
7
value. 10 New households’ indebtedness, which is secured on assets whose value is
increasing, the real estate assets in this case, displays lower interest rates, and requires
less additional guarantees and permits higher equity withdrawal. The increasing value of
the homeowners’ collateral during the pre-crisis period and the related ‘wealth’ effect,
contribute to boost the demand for credit not merely via residential investment. 11 It also
has a remarkable effect on other macroeconomic variables, which are a source of
demand for credit too, i.e. consumption, as highlighted by Bridges et al. (2006), Arestis
and Karakitsos (2008) and Case et al. (2011). We may also note that there is a loop
between credit and housing prices. In particular the possibility of obtaining cheaper and
higher volume of credit makes the ownership of dwelling more affordable and
attractive, which increases the demand for housing. This increase in demand promotes
housing price appreciation, since the supply of housing is fixed in the short run, which
strengths the ‘collateral’ channel and fuels the loop credit-housing prices. 12 The
inclusion of housing prices as a key determinant of credit demand is also justified
following Iacoviello and Minetti (2008). The latter considers housing prices as the most
informative element in terms of the evolution of housing demand in the short run, due to
the nature of the housing market and the prevalence of price adjustments.
Our contribution also accounts for the positive impact of real disposable income,
RDY, on the development of credit. Specifically, borrowers’ income is used by the
lenders to evaluate borrowers’ risk of default, since in a context of imperfect
information and uncertainty this is the basic indicator that commercial banks can utilize
to approximate households’ cash-flows.
13
This element influences the final risk
premium that the borrowers have to face and also affects other conditions of the
mortgage, as for example, its maturity, its volume or the requirement of additional
assets to secure the operation. Regarding total income, Almeida et al. (2006) points to
the so-called ‘income (or affordability) constraint’, i.e. households are finance10
See Benito et al. (2006), Corrado (2007) and Aron et al. (2012) for further theoretical explanations and
empirical analyses of the ‘collateral’ channel.
11
Miller et al. (2011) and Campbell and Cocco (2007) distinguish between the effects of unpredictable
and predictable variations in housing prices. The former influences the macroeconomy via the ‘wealth’
effect of households, while the latter proxies the ‘collateral’ channel. The influence of ‘collateral’ channel
becomes stronger when households are financially constrained. Miller et al. (2011) found that the impact
of the ‘collateral’ channel is stronger than ‘wealth’ effects in the US.
12
Although the causality between credit and housing prices goes in both directions, the strongest effect is
the one that emerges from housing prices to credit (Hofman, 2003; Goodhart et al., 2006).
13
The effect of disposable income on credit is uncertain (Nobili and Zollino, 2012). Specifically, rising
incomes induce individuals to go for higher indebtedness. Nevertheless, the demand for credit is not very
sensitive to variations in income in those markets where the loan-to-value ratio is low (see also Almeida
et al., 2006).
8
constrained in the sense that when they obtain a loan the annual total repayment of the
mortgage has to be a proportion of their total yearly income. 14 In terms of our model,
real disposable income is the main determinant of the demand for housing, and as a
result, we expect a relevant role in the explanation of the demand for credit. 15
The third determinant of the banking credit that our model considers is real
residential investment, RRI, which approximates the flow of new constructions. 16 Our
model assumes a positive and strong effect of residential investment on the demand for
banking finance, since an increase in the volume of acquisitions of real estate assets
drives up house prices in the short-run, which means that households have to face
higher debts to purchase new properties. However, in the long-run increasing real
residential investment means an increase in the supply of housing, which induces a
decline in housing prices and makes the asset more affordable. This improvement of the
affordability permits the entrance in the market of some households that were unable to
purchase a property before. This increase in the demand for housing boosts the demand
for credit, although the quantity that households have to borrow is lower than in the
short-run; there are, actually, more purchasers willing to get into debt. After some
period in which housing demand is growing, housing prices go up and a new cycle in
the housing market emerges; in this evolution of banking, debt materialises. 17
Moreover, investment in housing assets has an indirect effect on credit demand, which
reinforces the previous impact and is derived from an increase in the demand for
housing. Residential investment has strong ‘pulling’ effects on the economy in terms of
unemployment. 18 In this sense, an increase in the activity of the residential sector means
creation of income, which is distributed to households and makes more affordable this
kind of investment.
14
The impact on prices which emanates from changes in income is enlarged through fluctuations in the
demand for credit (Stein, 1995; Almeida et al., 2006).
15
Rising incomes reinforces the positive impact of housing prices on households’ indebtedness, as has
been demonstrated by Haurin and Rosenthal (2006) for the United States.
16
The model assumes that the dynamics in the housing market are demand-led, although in the short run
the supply of housing is given and all the pressures coming from the demand side have an impact on
prices, and eventually on the value of the collateral.
17
An alternative view of real residential investment considers that this variable is driven by the evolution
of Tobin’s q (Brainard and Tobin, 1977), which suggests that home builders develop their projects in
those situations where the price of the asset is higher than their cost of production, i.e. the Tobin’s q ratio
is above unity. This theory can be reconciled with our perspective, under the assumption that
extraordinary profitability, which emerges in the housing market due to an exogenous shock in incomes
or monetary policy, fuels the demand for housing. See Girouard and Blöndal (2001) for more discussion
of the Tobin’s q ratio in the housing market.
18
See Girouard and Blöndal (2001) and Iacoviello (2004) for a discussion of the role of this variable in an
analysis of consumption, credit and the housing market.
9
Finally, our theoretical proposition accounts for the relationship between the
mortgage rate, MR, and credit. The importance of this variable is multiple: (a) it is a
proxy for the user cost of dwelling, which influences the demand for housing; (b) it
accounts for the cost of external finance that households have to consider in order to
purchase a dwelling, which also exerts an effect on the demand for credit; (c) it is also
important in the supply side of the housing market, since some home builders have to
borrow funds to develop their activities; 19 and (d) it is the main explanatory variable of
the supply of credit. By considering the ratio of interest rate payments on consumer debt
over nominal disposable income as an indicator of affordability, we can expect a
slowdown of households’ credit demand when the price of mortgages increases, since
this means facing higher debt service burdens in those cases where disposable income is
constant. 20 Moreover, rising mortgage rates provoke contractions in the demand for
credit due to the fact that there is a fraction of the potential borrowers, which is not
solvent under this condition. 21 The inclusion of this element could register possible
credit crunch episodes; i.e. a sharp contraction of available credit could emerge if the
Central Bank tightened strongly credit standards and pushed up interest rates in
response to increasing risk of the borrowers’ insolvency. 22
3. Empirical Investigation
In this section the econometric technique, the dataset and our empirical results are
presented.
3.1 Econometric Technique
We begin by assuming a linear specification of equation (3) and applying the standard
cointegration technique (Engle and Granger, 1987). We deal with the estimation of the
19
In this sense a high mortgage rate could crowd out some construction firms, which would have a
negative impact on the supply of housing. As a result, acceleration of housing prices could take place,
whose effects on credit can be interpreted as described in the text.
20
See Aoki et al. (2002), Calza et al. (2007) and Corrado (2007) for a relevant discussion and empirical
evidence on the positive effects of rising house prices on the access and the cost of credit. More
specifically, the first study focuses on the US market, the second one considers some OECD countries
and the third study concentrates on the UK.
21
Under such a case, some households would have to opt to rent the dwelling services instead of
acquiring the property of the asset, which in the long run means a final negative effect on house prices;
this would induce a decline in the value of the collateral. This decline means a slowdown of credit
demand, which reinforces the direct negative effect that emerges from an increase in interest rates.
22
See Bernanke et al. (1991), Wolfson (1996) and Buera and Moll (2012) for further discussion and
alternative views on the implications of a credit crunch.
10
relationship under examination as proposed by Hendry and Nielsen (2007). The first
step is to estimate the long-run equilibrium relationship by means of Ordinary Least
Squares (OLS). Then the existence of cointegration among the variables is tested by
checking the stationarity of the residuals produced by the long-run relationship. We
utilize the augmented Dickey-Fuller (Dickey and Fuller; 1979, 1981) and the
Kwiatkowski-Phillips-Schmidt-Shin (Kwiatkowski et al., 1992) tests in our attempt to
examine whether the residuals are integrated of zero degree. In the second step, we
account for the short-run dynamics by estimating a model in differences, which also
includes an error-correction term. The latter variable, which is built as the laggedresiduals term of the cointegrating long-run relationship, shows the percentage of
disequilibria eliminated between the short-run and the long-run model in each period.
The short-run relationships are estimated by applying the ‘general to specific’ modelling
strategy (Hendry and Richard, 1983). 23
The order of integration of our dataset is studied by means of several unit
root/stationarity tests. We apply the augmented Dickey-Fuller (Dickey and Fuller; 1979,
1981), the GLS-based Dickey-Fuller (Nelson and Plosser, 1982) and the Phillips-Perron
(Phillips and Perron, 1988) tests, whose null hypothesis is the existence of a unit root in
the
variable
under
consideration.
The
Kwiatkowski-Phillips-Schmidt-Shin
(Kwiatkowski et al., 1992) test is also utilized to study the possible stationarity of the
data, since under the presence of structural breaks the results of the unit root/stationarity
tests could lead to the wrong order of integration. The use of cointegration methods in
order to avoid the problem of spurious regressions can be justified to the extent our data
are I(1), i.e. they present a unit root, as we try to verify empirically in what follows
below, where the mentioned unit root/stationarity tests are utilized.
We also check for the existence of cointegration by means of the Johansen’s
(1988, 1991) maximum eigenvalue and trace tests for cointegration to assure the
robustness of our results. Finally, the validity of the cointegrating relationships
estimated by OLS is also tested by means of the R-squared, the DW statistic, the Akaike
Information Criterion (AIC), the Schwartz Information Criterion (SIC) and the F-
23
A general model is initially estimated, which includes several lags of the endogenous and the
exogenous variables. Progressively, those statistically-insignificant variables are eliminated until the point
when a parsimonious model with all its significant determinants is reached.
11
statistic. 24 We also report other diagnostic/statistics, which are utilized in the event of
the short-run estimations: (a) the Breusch-Godfrey Serial Correlation LM (Breusch,
1979; Godfrey, 1978) test, which detects the existence of autocorrelation of first-,
second- and third-order; (b) the White (White, 1980) test, with and without cross terms,
which checks for homoscedasticity; and (c) the ARCH (Engle, 1988) test, which
permits to test for the absence of ARCH effects of first- and second-order in the
residuals.
We estimate a standard semi-log-linear model as shown in equation (9):
BC = γ 0 + γ 1 PH + γ 2 RDY + γ 3 RRI − γ 4 MR
(9)
where the meaning of variables is as explained above. All the variables are expressed in
terms of logarithms except the mortgage rate.
3.2 Data
Our theoretical model is tested by means of samples for various countries, which
includes annual observations from 1970 till 2011 for the following 9 OECD countries:
Belgium, Finland, Italy, Japan, the Netherlands, New Zealand, Norway, Spain and the
United States. The length of the period under consideration is determined by the
availability of historical information in the sources consulted. The size of our sample is
rich enough to study the relationship between the housing market and the bank credit in
the private sector in countries with important differences in terms of development and
structure of the financial sector and different role of the housing market.
The endogenous variable of our model, the volume of bank credit, is
approximated by the Domestic Credit to the Private Sector (as a Percentage of GDP)
time series, which are available on the World Bank databank. 25
An important source of information is the AMECO databank, which is published
by the European Commission’s Directorate General for Economic and Financial
24
The AIC and SIC are useful in selecting the model, which fits better to the features of the data under
consideration when there are several specifications for the same relationship. In particular, the chosen
model is the one, which exhibits the lowest values for both criteria (Gujarati, 1997).
25
These series are available at:
http://data.worldbank.org/
12
Affairs. 26 The following statistics are obtained: (a) Gross Fixed Capital Formation by
type of Goods at Current Prices (Dwelling); (b) Gross National Disposable Income per
Head of Population; (c) Gross Domestic Product Price Deflator; and (d) Real Long-term
Interest Rate. 27 The lack of homogeneous and reliable information for mortgages rates
during the period under consideration forces us to use the long-term interest rate its
proxy.
The Bank of International Settlements (BIS) is also consulted to obtain the Real
House Prices Index, which is available for the period 1970-2011. 28
E-Views 5.0 is the econometric package utilized to run the models and those
tests, which are used for the purposes of our study.
3.3 Empirical Results
We check for the presence of unit roots in our data by applying several unit
root/stationarity tests. We accept the null hypothesis of the presence of unit roots in the
case of the augmented Dickey-Fuller (Dickey and Fuller; 1979, 1981), the PhillipsPerron (Phillips and Perron, 1988) and the GLS-based Dickey-Fuller (Nelson and
Plosser, 1982) tests. These findings are supported by the rejection of the null hypothesis
of stationarity in the case of the Kwiatkowski-Phillips-Schmidt-Shin test (Kwiatkowski
et al., 1992). These results lead to the conclusion that our data are of first order
integration, i.e. I(1). 29
3.3.1 Long-run Equilibrium Relationships
The empirical relationships, which are modelled to analyse the long run are displayed in
Table 1A. 30 The statistics applied to validate the econometric regressions are reported in
26
This information is available at:
http://ec.europa.eu/economy_finance/db_indicators/ameco/index_en.htm
27
The OECD databank is used to complete some missing information. We utilize Gross fixed capital
formation and housing to replace the missing values in dwelling series for Norway, and the Long-term
interest rate in the case of Norway and New Zealand.
28
This data is available at:
http://www.bis.org/
29
The results of all these unit root/stationarity tests are not displayed in the paper but are available from
the authors upon request.
30
The Johansen’s (1988, 1991) test suggests cointegration at 5% significance level in all the countries
except in Japan, where a cointegrating relationship is found at the 10% significance level. The results of
this test are not reported but are available from the authors upon request.
13
Table 1B. 31 All the cointegrating relationships include an intercept which is negative
and significant in all the cases.
TABLE 1A CREDIT LONG-RUN RELATIONSHIP (1970-2011).
Long-run relationship
Constant
L_RHP
L_RDY
L_RRI
MR
1.080865*** (0)
-3.291772*** (0)
Belgium
-8.816358***
2.665025*** (0)
Finland
-3.950581***
Italy
-4.572499***
Japan
-7.356197***
Netherlands
-3.115259***
1.080747*** (0)
New Zealand
-2.742257***
1.457735*** (0)
Norway
-6.519227***
Spain
-2.945305***
US
-5.557265***
0.764384*** (0)
0.27513*** (0)
0.814936*** (0)
1.075322*** (0)
0.648469*** (0)
-1.617144** (0)
1.712545*** (0)
Note: ***, ** and * indicate statistical significance and rejection of the null at the 1, 5 and
10 percent significance levels, respectively. Numbers in parentheses, in the case of the variables,
show the lag(s) of the relevant variable.
TABLE 1B CREDIT LONG-RUN RELATIONSHIP DIAGNOSTICS/STATISTICS I.
Diagnostic/Statistics Long-run Relationship
R-squared
DW
AIC
SIC
F-statistics
Belgium
0.920258
0.596177
-0.715964
-0.632375
450.0795 (0.0000)
Finland
0.523184
0.205051
-0.503058
-0.419469
42.79252 (0.0000)
Italy
0.745851
0.635862
-1.205774
-1.08039
55.75922 (0.0000)
Japan
0.921629
0.709889
-2.764604
-2.63922
223.4361 (0.0000)
Netherlands
0.843016
0.421676
-0.82782
-0.744231
209.4329 (0.0000)
New Zealand
0.734979
0.292691
1.309284
1.392873
108.158 (0.0000)
Norway
0.740873
0.208035
-0.517707
-0.430631
100.0691 (0.0000)
Spain
0.732446
0.214965
-0.470847
-0.345464
52.01382 (0.0000)
US
0.935325
0.55168
-2.308628
-2.225039
564.0178 (0.0000)
Note: In the last column numbers in parentheses indicates the p-value of each test.
Table 1A reports the direct relationship between housing prices and the volume
of bank credit. Specifically, the strength of the ‘collateral’ channel is remarkable in
those countries, which suffer a devastating bubble in the housing market, as for
example, Spain (0.648) and Finland (0.764). Interestingly enough, in 2008 housing
prices in Spain were 24% above their fundamentals, while in Finland they were
overvalued by 15%. These two countries showed then the strongest deviation between
housing prices and their fundamentals in the Euro Area (El Mundo, 2010). The lowest
31
The robustness of these models is checked by means of the OLS White-Heteroskedasticity Consistent
technique (White, 1980), which corrects for the presence of heterokedasticity; and the Generalised
Method of Moments (Arellano and Bover, 1995), which deals with autocorrelation of unknown forms and
simultaneity problems. These estimations are not shown since their results in terms of significance and
value of the coefficients are similar to those obtained by OLS reported in Table 1A.
14
incidence of housing prices on demand for credit emerges in the Japanese market
(0.275).
Disposable income is the only variable, which drives demand for credit, in the
case of Belgium (2.665), the United States (1.713) and Norway (1.075). It is also
significant in Japan (0.815), although the effect is lower in comparison with the one,
which is found in Belgium and the United States.
The flow of construction is relevant in the cases of New Zealand (1.458), the
Netherlands (1.081) and Italy (1.081). In the first two markets, this variable, which
comes in our model through the supply side of the housing market, is the only factor
that determines the long-run equilibrium. 32
Our study also points to a negative effect of the cost of external finance in the
two Southern European economies that we consider, i.e. Italy and Spain. This last effect
(-1.617) is around half of the size of the one found in the Italian economy (-3.292).
We discuss next the diagnostics/statistics applied in order to validate our models.
In general terms, the adjustment of the models is high, since the R-squared is higher
than 70% in all the econometric models except in Finland, where the model is able to
explain just the 52% of the variation in bank credit. The second column reports the
value of the Durbin Watson statistics, which are not close to 2 since there is
autocorrelation. 33 We also report two information criteria (the AIC and the SIC). These
statistics are useful to select the specification that fits better to the structure of the data
when there are several alternatives to model the behaviour of the endogenous variable. 34
Table 1B shows the F-statistic where the rejection of the null hypothesis in all the cases
is relevant, which enables us to conclude on the joint significance of the variables
included in the models.
3.3.2 Short-run Dynamics
Table 2A displays the models, which are estimated to capture the dynamics of the
model in the short run. Tables 2B and 2C report the results of the statistics/diagnostics,
which were presented in section 3.1.
32
The strong impact in New Zealand is easily understood since in 2007 the housing market in this
economy was stagnated and after then prices continue growing up again.
33
According to the cointegration technique (Engle and Granger, 1987) the lack of autocorrelation has to
be checked just in the error-correction model.
34
Gujarati (1997) recommends choosing the model whose absolute values of the AIC and SIC are the
lowest possible.
15
TABLE 2A CREDIT SHORT-RUN RELATIONSHIP (1970-2011).
Short-run Relationship
Constant
∆L_RHP
∆L_RDY
∆L_RRI
∆MR
∆L_CREDIT
EL_CREDIT
Belgium
0.028389
Finland
0.008079
0.267139* (3)
Italy
0.002945
0.708952*** (1)
-0.138908*
Japan
0.002373
0.477998*** (1)
-0.350294***
0.23186*** (2)
-0.265906**
0.496211*** (1)
-0.073486*
Netherlands
0.036917***
0.172884** (0)
-0.103185**
New Zealand
0.034756***
0.210688*** (0)
-0.035089**
0.312063** (1)
Norway
0.011056
Spain
0.007648
US
0.012611*
0.389465** (1)
-0.104594**
0.707276*** (1)
-0.089730**
0.522043* (0)
-0.198502**
Note: ***, ** and * indicate statistical significance and rejection of the null at the 1, 5 and 10 percent
significance levels, respectively. Numbers in parentheses, in the case of the variables, show the lag(s)
of the relevant variable.
TABLE 2B CREDIT SHORT-RUN RELATIONSHIP DIAGNOSTICS/STATISTICS I.
Diagnostic/Statistics Short-run Relationship
R-squared
DW
AIC
SIC
F-statistics
Belgium
0.237075
2.033039
-1.457911
-1.327296
5.282674 (0.0101)
Finland
0.668652
2.322958
-3.648111
-3.475734
22.87036 (0.0000)
Italy
0.401239
2.111268
-3.360975
-3.233009
12.06208 (0.0001)
Japan
0.348626
2.082007
-3.767273
-3.639307
9.633914 (0.0004)
Netherlands
0.783877
1.985935
-3.643707
-3.474819
43.52391 (0.0000)
New Zealand
0.908671
1.603619
-3.263641
-3.094753
119.3928 (0.0000)
Norway
0.5112
1.956996
-3.148235
-2.970481
10.80689 (0.0000)
Spain
0.494425
1.717818
-3.304016
-3.176050
17.60305 (0.0000)
US
0.13882
2.417253
-3.476376
-3.349711
2.982144 (0.0630)
Note: In the last two columns numbers in parentheses indicates the p-value of each test.
TABLE 2C CREDIT SHORT-RUN RELATIONSHIP DIAGNOSTICS/STATISTICS II.
Diagnostic/Statistics Short-run Relationship
LM (1)
LM (2)
LM (3)
White
Belgium
0.025450 (0.874223)
Finland
1.74714 (0.195327)
0.048645 (0.952590)
0.099307 (0.959800)
1.991267 (0.119564)
0.871257 (0.428115)
0.809402 (0.498349)
0.485415 (0.814063)
Italy
1.707244 (0.199864)
1.300146 (0.285696)
1.07211 (0.374284)
0.226865 (0.921461)
Japan
1.82672 (0.185187)
1.526363 (0.231846)
1.395053 (0.261611)
Netherlands
0.363828 (0.550277)
0.53111 (0.592750)
0.514325 (0.675257)
New Zealand
1.45194 (0.236301)
1.783492 (0.183415)
Norway
0.02439 (0.876942)
0.609972 (0.550190)
Spain
0.016358 (0.898963)
US
3.320754 (0.076726)
White X
ARCH (1)
ARCH (2)
1.781107 (0.145946)
0.001304 (0.971407)
0.003147 (0.996859)
0.335323 (0.955347)
0.014571 (0.904610)
0.12686 (0.881285)
0.192652 (0.963225)
0.051246 (0.822189)
0.083401 (0.920170)
1.186009 (0.334527)
1.087558 (0.385486)
0.752416 (0.391458)
0.619951 (0.543941)
0.323178 (0.895520)
0.28199 (0.941382)
1.672664 (0.203919)
1.116607 (0.338771)
1.160444 (0.339545)
1.860725 (0.127342)
1.618218 (0.173245)
0.035461 (0.851663)
0.200356 (0.819372)
0.394522 (0.757911)
0.385388 (0.882088)
0.386023 (0.930674)
0.044034 (0.835121)
0.048124 (0.953089)
0.146955 (0.863878)
0.335906 (0.961949)
0.378085 (0.822664)
0.307265 (0.905035)
6.470822 (0.015403)
0.474490 (0.626261)
1.843544 (0.173295)
1.367418 (0.269259)
1.819876 (0.147026)
2.212998 (0.075659)
2.187768 (0.147575)
2.056208 (0.143115)
Note: Numbers in parentheses indicates the p-value of each test.
The estimations of the dynamics in the short run reveal that the mortgage rate
has no impact in this time horizon. This finding can be interpreted according to the Post
Keynesian view, which assumes that the demand for credit is inelastic to fluctuations in
the rate of interest (Lavoie, 1984).
16
The models estimated to capture the short-run effects include a constant, which
is positive and significant in the cases of the Netherlands, New Zealand and the United
States. Our estimations also include a lag of the endogenous variable, the bank credit.
This element drives the evolution of credit in this time horizon in the credit markets of
Spain, Italy and Japan. The strongest impact is found in Spain (0.707) and Italy (0.709).
The coefficient, which emerges from the Japanese economy, is lower than in the
markets just considered (0.478). In Finland this parameter is slightly superior (0.496).
The lowest impact emerges in the Norwegian case (0.389).
Our empirical analysis indicates how rising housing prices exert a positive effect
on credit in the half of the countries under consideration. Specifically, Finland (0.232)
shares this element as the key explanatory factor of bank credit in the short and long
run. The highest impact is found in Norway (0.312), while the lowest effect emerges in
the Netherlands (0.173) and New Zealand (0.211).
Residential investment contributes to explaining the development of credit just
in Belgium (0.267). This result conforms to our theoretical hypothesis, which assumes
that the relationship between both variables is positive.
Our estimations also find positive effect of the disposable income variable only
in the case of the United States (0.522). As in the long run, this variable is the factor that
drives the evolution of bank lending in the United States economy.
The estimated models include an error-correction term which is negative and
significant. In all the cases except in New Zealand (-0.035), Finland (-0.073) and Spain
(-0.090), the percentage of adjustment is higher than 10%. The speed of adjustment is
around 10% in Norway and the Netherlands. Slightly more dynamic is the Italian
market where the error-correction term reduces to a 14% of the disparities every year.
The highest values for this term appear in Japan and Belgium, where the model
eliminates a 35% and 27% of the disequilibria in each period respectively. The United
States also displays dynamic markets where the error-correction term is almost 20%.
In order to check the validity of our results we begin with the discussion of the
R-squared. This value is very high in the case of New Zealand (91%) and the
Netherlands (78%). In contrast, the United States model is able to explain only 14% of
the fluctuations. The adjustment of the model is also high in Finland (66%), Norway
(51%) and Spain (49%). Table 2B also reports the Durbin Watson statistic, whose
values are around 2 in all the cases, except in Ireland and the United States where the
values are close to 1.5 and 2.5 respectively. The AIC and the SIC, which are reported in
17
the third and fourth columns, show negative values. The models displayed in Table 2A
were chosen according to both criteria, by selecting those with the lowest value for this
statistic. The last column of this table presents the F-statistics, which conclude the joint
significance of the variables included in these estimations.
We check the lack of autocorrelation of first-, second- and third-order by using
the Breusch-Godfrey Serial Correlation LM test. In all the models we accept the null
hypothesis of absence of autocorrelation. Table 2C also reports the values of the White
test (without and with cross terms), which show homoscedasticity in the residuals of the
models. The ARCH tests state the absence of ARCH effects of first- and second-order.
In general terms, all the models satisfy the tests, which were applied at the 5%
significance level. 35
4. Overall Discussion of the Theoretical and Empirical Results
The econometric exercise, which is presented in the previous section, reinforces
empirically our theoretical frame. The assumption of a monetary production economy,
where money-credit is created by the commercial banks in response to the demand for
credit, permits us to endogenise bank lending by considering households’ investment in
housing as the key element that induces households to get into debt.
Our analysis confirms the evolution of housing prices as the most important
variable in the explanation of bank credit. This fact highlights the importance of the
‘collateral’ channel and the positive impact on demand for housing, which emanates
from the ‘wealth’ effect that households experiment when there is an increase in the
value of their assets. Incomes are also a relevant factor in the explanation of demand for
housing and mortgages. Its influence is much more intense in the long run rather than in
the short run, and it is particularly strong in the case of Belgium and the United States.
These findings are perfectly understandable, since the demand for credit is driven by the
activity in the housing market whose main determinants are those, which define the
housing affordability.
Our estimations also capture the role of residential investment and mortgage
rates, whose effects take place mainly in the long run. The cost of external finance,
which approximates the user cost of dwelling, is an explanatory variable in Spain and
Italy, where the impact is twice the one in the Spanish market. However, this variable
35
In the case of Spain we relax the level of significance to 1% to accept the absence of ARCH effect of
first order.
18
does not exert a persistent influence, which can be justified according to our theoretical
framework, where the mortgage rate enters the model through the existence of a
horizontal supply for credit. The flow of new residential assets, which are produced in
response to demand for housing, is also included. In particular, residential investment,
whose influence in the equilibrium of the housing market comes from the supply side of
the market, exerts a positive effect on demand for credit since this investment requires
strong external finance.
5. Summary and Conclusions
The aim of this contribution is to endogenize the development of bank credit
following the notion of endogenous money-credit. We assume a dynamic monetized
production economy where the main source of demand for credit emerges from
households’ desires to acquire housing. Our results conclude that bank credit is fuelled
by housing prices, residential investment and disposable income. Changes in the
mortgage rate affect the demand for credit negatively, since this means a significant
change in the user cost of dwelling, which provokes a change in the demand for
housing. The most important variables in the determination of demand for credit are
those which determine the affordability of housing, i.e. prices and incomes.
In terms of economic policy, the most important tool that monetary authorities
should utilize in order to avoid severe problems in the financial system is prudential
policy. This is much more useful than actions via interest rates, since the role of the
commercial banking sector is to provide the liquidity, which is required to develop
economic activities. In this context, it is much more important for commercial banks to
identify those households, which are credit-worthy rather than manipulate interest rates
or focus on the compensations that they could obtain in case of households’ default.
Policymakers should also take into account the evolution of asset prices, since the
conditions and the size of mortgage provision are related to the value of the asset, which
is used to back the relevant loans.
References
Aoki, K, Proudman, J. and Vlieghe, G. (2002), “Houses as a Collateral: Has the Link
between House Prices and Consumption in the U.K. Changed?”, FRBNY Economic
Policy Review, 8 (1): 163-167.
19
Adelino, M., Schoar, A. and Severino, F. (2012), “Credit Supply and House Prices:
Evidence from Mortgage Market Segmentation”, NBER Working Paper No. 17832,
Cambridge MA: National Bureau of Economic Research.
Almeida, H., Campello, M. and Liu, C. (2006), “The Financial Accelerator: Evidence
from International Housing Markets”, Review of Finance, 10 (3): 1-32.
AMECO (2011), AMECO Databank List of Variables. Available at:
http://ec.europa.eu/economy_finance/db_indicators/ameco/index_en.htm
Arellano, M. and Bover, O. (1995), “Another Look at the Instrumental Variable
Estimation of Error-components Models”, Journal of Econometrics, 68 (1): 29-51.
Arestis, P. and Karakitsos, E. (2008), “The US Housing Slump and the Consumer”,
Journal of Post Keynesian Economics, 30 (3): 335-352.
Aron, J., Duca, J.V., Murata, K., and Murphy, A. (2012), “Credit, Housing Collateral,
and Consumption: Evidence from Japan, the U.K. and the U.S.”, Review of Income and
Wealth, 58 (3): 397-423.
Benito, A., Thompson, J., Waldron, M. and Wood, R. (2006), “House Prices and
Consumer Spending”, Quarterly Bulletin, Summer, London: Bank of England.
Bernanke, B.S. and Gertler, M. (1989): “Agency Cost, Net Worth, and Business
Fluctuations”, American Economic Review, 79 (1): 3-55.
Bernanke, B.S. Lown, C.S. and Friedman, B.M. (1991), “The Credit Crunch”,
Brookings Papers on Economic Activity, 1991 (2): 205-247.
Bernanke, B.S., Gertler, M. and Gilchrist, S. (1999), “The Financial Accelerator in a
Quantitative Business Cycle Framework”, in Handbook of Macroeconomics, Taylor,
J.B. and Woodford, M. (eds.), Vol. I, Amsterdam: Elsevier Science.
Bernanke, B.S. (2007), “The Financial Accelerator and the Credit Channel”, Speech
15th June. Available at:
http://www.federalreserve.gov/newsevents/speech/bernanke20070615a.htm
Brainard, W.C. and Tobin, J. (1977), “Asset Markets and the Cost of Capital”, in
Economic Progress, Private Values and Public Policy: Essays in Honour of William
Fellner, Belassa, B. and Nelson, R. (eds.), Vol. I, New York: North-Holland.
Breusch, T.S. (1979), “Testing for Autocorrelation in Dynamic Linear Models”,
Australian Economic Papers, 17 (31): 334–355.
Bridges, S., Disney, R. and Gathergood, J. (2006), “Housing Collateral and Households
Indebtedness: Is There a Household Financial Accelerator”, in The Economics of
Consumer Credit, Bertola, G., Disney, R. and Grant, C. (eds.), Cambridge MA: MIT
Press Cambridge.
Buera, F.J. and Moll, B. (2012), “Aggregate Implications of a Credit Crunch”, NBER
Working Paper No. 17775, Cambridge MA: National Bureau of Economic Research.
20
Calza, A., Monacelli, T., and Stracca, L. (2007), “Mortgage Markets, Collateral
Constraints and Monetary Policy: Do Institutional Factors Matter ?”, CFS Working
Paper Series No. 2007/10, Frankfurt: Centre for Financial Studies.
Campbell, J.Y. and Cocco, J.F. (2007), “How Do House Prices Affect Consumption?
Evidence From Micro Data”, Quarterly Journal of Economics, 118 (4): 1449-1494.
Case, K.E., Quigley, J.M. and Shiller, R.J. (2011), “Wealth Effects Revisited 19782009”, NBER Working Paper No. 16848, Cambridge MA: National Bureau of
Economic Research.
Corrado, G. (2007), “Modelling Housing and Consumers’ Spending under Liquidity and
Borrowing Constraints: Some Stylised Facts from UK Housing Market”, Ekonomia, 10
(2): 112-137.
Daugherty, M.R. (1942), “The Currency-Banking Controversy: Part 1”, Southern
Economic Journal, 9 (2): 140-155.
Dickey, D. A. and Fuller, W.A. (1979), “Distribution of the Estimator for
Autoregressive Time Series with a Unit Root”, Journal of the American Statistical
Association, 74 (366): 427-431.
Dickey, D.A. and Fuller, W.A. (1981), “Likelihood Ratio Statistics for Autoregresssive
Time Series with a Unit Root”, Econometrica, 49 (4): 1057-1072.
El Mundo (2010), La Vivienda Española Aún Sigue Sobrevalorada un 17%. 5 October.
Engle, R.F. and Granger, C.W.J. (1987), “Co-integration and Error Correction:
Representation, Estimation and Testing”, Econometrica, 55 (2): 251-276.
Engle, R.F. (1988), “Autoregressive Conditional Heteroscedasticity with Estimates of
the Variance of United Kingdom Inflation”, Econometrica, 96 (5): 893-920.
Girouard, N. and Blöndal, S. (2001), “House Prices and Economic Activity”, OECD
Economics Department Working Papers No. 279, Paris: OECD Publishing.
Godfrey, L.G. (1978), “Testing Against General Autoregressive and Moving Average
Error Models when the Regressors Include Lagged Dependent Variables”,
Econometrica, 46 (6): 1293–1302.
Goodhart, C. (1995), “Price Stability and Financial Fragility”, in Financial Stability in a
Changing Environment, Sawamoto, K., Nakajima, Z. and Taguchi, H. (eds.), London:
Macmillan.
Goodhart, C., Hofmann, B. and Segoviano, M. (2006), “Default, Credit Growth and
Assets Prices”, IMF Working Paper No. 223, Washington, D.C.: International Monetary
Fund.
Graziani, A. (2003), The Monetary Theory of Production, Cambridge: Cambridge
University Press.
21
Gujarati, D. (1997), Econometría Básica, Bogotá: McGraw-Hill Interamericana.
Haurin, D.R. and Rosenthal, S.S. (2006), “House Prices Appreciation, Savings and
Consumer Expenditures”, Ohio State University Working Paper, Columbus OH: Ohio
State University.
Hendry, D. and Richard, J. (1983), “The Econometric Analysis of Economic Time
Series”, International Statistics Review, 51 (2): 3-33.
Hendry, D. and Nielsen, B. (2007), Econometric Modeling, Princenton: Princenton
University Press.
Hofmann, B. (2003), “Bank Lending and Property Prices: Some International
Evidence”, HKIMR Working Paper No. 22/2, Hong Kong: Hong Kong Institute for
Monetary Research.
Hofmann, B. (2004), “The determinants of Private Sector Credit in Industrialised
Countries: Do Property Prices Matter?”, International Finance, 7 (2): 203-234.
Iacoviello, M. and Minetti, R. (2008), “The Credit Channel of Monetary Policy:
Evidence from the Housing Market”, Journal of Macroeconomics, 30 (1): 69-96.
Iacoviello, M. (2004), “Consumption, House Prices, and Collateral Constraints: A
Structural Econometric Analysis”, Journal of Housing Economics, 13 (4): 304-320.
Johansen, S. (1988), “Statistical Analysis of Cointegrating Vectors”, Journal of
Economic Dynamics and Control, 12 (2-3): 231-254.
Johansen, S. (1991), “Estimation and Hypothesis Testing of Cointegration Vectors in
Gaussian Vector Autoregressive Models”, Econometrica, 59 (6): 1551–1580.
Kaldor, N. (1982), The Scourge of Monetarism. Oxford: Oxford University Press.
Kiyotaki, N. and Moore, J. (1997), “Credit Cycles”, Journal of Political Economy, 105
(2): 211-248.
Kwiatkowski, D., Phillips, P.C.B., Schmidt, P. and Shin, Y. (1992), “Testing the Null
Hypothesis of Stationarity against the Alternative of a Unit Root: How Sure Are We
that Economic Time Series Have a Unit Root?”, Journal of Econometrics, 54 (1-3):
159-178.
Lavoie, M. (1984), “The Endogenous Flow of Credit and the Post Keynesian Theory of
Money”, Journal of Economics Issues, 18 (3): 771-797.
Miller, N., Peng, L. and Sklarz, M. (2011), “House Prices and Economic Growth”,
Journal of Real Estate Finance and Economics, 42 (2): 522-541.
Moore, B.J. (1983), “Unpacking the Post Keynesian Black Box: Bank Lending and the
Money supply”, Journal of Post Keynesian Economics, 5 (4): 537-57.
22
Nelson, C.R. and Plosser, C. (1982), “Trends and Random Walks in Macroeconmic
Time Series: Some Evidence and Implications”, Journal of Monetary Economics, 10
(2): 139-162.
Nobili, A. and Zollino, F. (2012), “A Structural Model for the Housing and Credit
Market in Italy”, Banca d’Italia Working Papers No. 887, Rome: Banca d’Italia.
Phillips, P.C.B. and Perron, P. (1988), “Testing for a Unit Root in Time Series
Regression”, Biometrika, 75 (2): 335-346.
Stein, J. (1995), “Prices and trading volume in the housing market: A model with down
payment effects”, Quarterly Journal of Economics, 110 (2): 379-406.
White, H. (1980), “A Heteroskedasticity-Consistent Covariance Matrix Estimator and a
Direct Test for Heteroskedasticity”, Econometrica, 48 (4): 817–838.
Wolfson, M.H. (1996), “A Post Keynesian Theory of Credit Rationing”, Journal of Post
Keynesian Economics, 18 (3): 443-470.